OMB CARES_Appendix C_Public Use File-Data Masking

Appendix C:  Public Use File - Data Masking

DHHS/ACF/OPRE Head Start

Classroom-based Approaches and

Resources for Emotion and Social skill

promotion (CARES) project:

Tracking Participants

Public Use File - Data Masking Procedures

Prepared by MDRC for the U.S. Department of Health and Human Services, Office of the Assistant Secretary for Planning and Evaluation under Task Order 4, Contract No. HHS-282-00-0014.

1. Data Masking Strategies

MDRC implemented similar data masking procedures for this public use file as were used when creating the public use files for DHHS-ASPE and DHHS-ACF from the main NEWWS study. These procedures make it impossible to identify sample members directly and extremely difficult to infer identity based on small cell sizes. See the accompanying table for details about procedures for specific measures.

1. Sample member identifiers and characteristics.

• The NEWWS LT PUF uses a randomly-generated ID number (LTPUFID) as the only sample member identifier. This ID number differs from the ID number that DHHS programmers created when saving data for NEWWS sample members extracted from NDNH and HPB.

• Random assignment date, month, and quarter were excluded from the file and replaced by Random Assignment Year.

• Sample members’ date of birth was excluded from the file; sample members’ age at random assignment was collapsed to six values: 18, 20, 25, 35, 45, and 60.

• Child ages excluded from the file. Categorical and 0/1 measures based on age ranges (e.g., Age of youngest child 0-2, 3-5, 6-11, 12-18 years) were retained.

• Location of random assignment: City- or county-level measure of Random Assignment Site (Atlanta, Columbus, …) was retained on the file, but local welfare-to-work office code was excluded from the file.

2. Date variables

• All date variables were excluded from the file.

• No variable names include dates (e.g., EARN1, EARN2… rather than EARN91Q2, EARN91Q3…)

3. Earnings measures from statewide UIW systems and from NDNH

• Earnings reported by individual employers were summed to quarterly totals.

• Quarterly earnings totals were rounded to the nearest $100 or to $50 if between $0.01 and $49.99.

• Employer ID numbers were excluded from the file.

• Quarterly earnings were top-coded at $15,000 for statewide UIW data and $19,500 for NDNH data.

• For the series of quarterly earnings totals that were converted to 2006 dollars, totals were rounded again after conversion.

• For NDNH data, FIPS state codes were excluded from the file. “Employed in random assignment state” and “Employed in other state” are the only locations retained.

4. AFDC/TANF and fool stamp measures from state and county systems

• Payments were summed to yearly totals. All monthly and quarterly measures were excluded from the file.

• Yearly totals were rounded to the nearest $100 or to $50 if between $0.01 and $49.99.

• Yearly totals were top-coded at $20,000.

• 0/1 indicators of receipt of benefits were aggregated to quarterly measures. All monthly indicators of receipt were excluded from the file.

5. TANF receipt measures from HPB

• HPB contains 0/1 indicators of receipt but no payment data.

• 0/1 indicators of receipt of benefits aggregated to quarterly measures. All monthly indicators of receipt were excluded from the file.

• FIPS state codes were excluded from the file. “Received TANF in random assignment state” and “Received TANF in other state” are the only locations retained.

6. UI Benefits measures from NDNH

• UI Benefits were summed to quarterly totals.

• Quarterly Benefits totals were rounded to the nearest $100 or to $50 if between $0.01 and $49.99.

• Quarterly Benefits totals were top-coded at $19,500.

• FIPS state codes were excluded from the file. “Received UIB in random assignment state” and “Received UIB in other state” are the only locations retained.

7. Responses to the Two-Year- and Five-Year Client Surveys

• Hours per week of employment measures collapsed to: 0, 1-19, 20-29, 30-34, 35-39, 40-72, 73 or more.

• Average hourly wage measures rounded to the nearest $1 and top-coded at $25.

• Average weekly wage measures rounded to the nearest $25 and top-coded at $1,150